Different Role of Tumor Necrosis Factor-α Polymorphism in Non-Hodgkin Lymphomas among Caucasian and Asian Populations: A Meta-Analysis

Tumor necrosis factor-α (TNF-α) is an immunoregulatory cytokine involved in B- and T-cell function, and also plays an important role in inflammation and cancer. TNF-α-308G>A has been associated with constitutively elevated TNF-α expression. Several studies have reported the association between the TNF-α-308G>A polymorphism and non-Hodgkin lymphomas (NHL) risk, however, results are still inconsistent. To solve these conflicts, we conducted the first meta-analysis to assess the effect of TNF-α-308G>A polymorphism on the risk of NHL and various subtypes (additive model) including 10,619 cases and 12,977 controls in Caucasian and Asian populations. Our meta-analysis indicated that TNF-α-308G>A polymorphism is not associated with NHL risk when pooling all studies together (OR = 1.06, 95% CI: 0.92–1.23, p = 0.413). In stratified analyses, we found TNF-α-308A allele was significantly associated with higher risk of NHL, B-cell lymphomas (BCL), T-cell lymphomas (TCL) and diffuse large B-cell lymphomas (DLBCL) in Caucasians (OR = 1.22, 95% CI: 1.06–1.40, p = 0.007; OR = 1.18, 95% CI: 1.03–1.34, p = 0.014; OR = 1.20, 95% CI: 1.01–1.42, p = 0.040; OR = 1.21, 95% CI: 1.11–1.32, p < 0.001, respectively). Interestingly, it was associated with decreased risk of NHL, BCL and DLBCL in Asians (OR = 0.75, 95% CI: 0.66–0.86, p < 0.001; OR = 0.70, 95% CI: 0.52–0.94, p = 0.018; OR = 0.70, 95% CI: 0.57–0.86, p = 0.001). These findings also suggest TNF-α might play a distinct role in pathogenesis of NHL in different populations.


Introduction
Non-Hodgkin lymphomas (NHL), a complex group of heterogeneous diseases of uncontrolled B-or T-cell proliferation with distinct clinical and histological features, accounts for approximately 90% of all malignancy lymphomas [1]. Malignant transformation of B-or T-cells can occur at different stages of maturation, which reflects the heterogeneity of malignancies with various biologic and clinical behaviors. B-cell lymphomas (BCL) comprise 90% of NHL. Diffuse large B-cell lymphomas (DLBCL) and follicular lymphomas (FL) are the two major subtypes of BCL. Clinical outcome of NHL varies from subtype, diagnosis and response to treatment, however, prognosis of T-cell lymphoma (TCL) is usually worse than that of BCL. Etiology of NHL is still poorly understood, although epidemiological studies have shown that individuals with innate or acquired immune deficiencies, immunosuppression and infection are at increased risk of NHL [2,3]. Recently, accumulating evidence has suggested that genetic variations such as single nucleotide polymorphisms (SNPs) are associated with NHL risk and survival [4][5][6][7][8][9]. Moreover, previous studies showing a 2-to 3-fold risk of NHL with a family history of hematological malignancies indicates that genetic factors might play a critical role in NHL pathogenesis [10][11][12].
Tumor necrosis factor-α (TNF-α) is one of the most important pro-inflammatory and tumor-related cytokines for its regulating immune response, inflammation, Th1/Th2 balance and lymphomagenesis [13].
Increased serum values of TNF-α have been detected in autoimmune disease and many malignancies including lymphomas [14][15][16][17]. TNF-α-308G>A (rs1800629) SNP has increased susceptibility to many kinds of tumors and autoimmune diseases, such as hepatocellular carcinoma, myeloma, lymphoma, ulcerative colitis, and Crohn's disease [18][19][20]. TNF-α-308A allele is associated with higher constitutive and inducible TNF-α expression by affecting a consensus binding site of a transcription factor named activator protein-2 (AP-2) [21,22]. Studies using knockout mouse have supported that this cytokine could affect progression of BCL directly or indirectly [23,24]. Although the TNF-α-308G>A polymorphism has been widely assessed in association with NHL in different ethnicities, due to various sample sizes and genotyping methods, possibly because of NHL heterogeneity and other reasons, the results are still controversial.
In this study, we conducted the first comprehensive meta-analysis to test whether the TNF-α-308 polymorphism is associated with NHL overall risk or its subtypes, especially BCL, TCL, DLBCL, FL, chronic lymphocytic leukemia/small lymphocytic lymphomas (CLL/SLL), mantel cell lymphomas (MCL), mucosal-associated lymphomas (MALT), peripheral T-cell lymphomas (PTCL) and natural killer/T-cell lymphomas (NK/TCL). We also performed subgroup analysis by descent (Caucasians and Asians) to assess a possible factor that might influence the overall results. Therefore, this study might have more statistical power and increase precision to estimate association between TNF-α-308 polymorphism and its effect on NHL.

Eligible Studies
In the initial screening for key words, 405 potential articles were identified in PubMed, Embase and Cochrane Library. After removing duplication, 321 articles were needed for further assessment. Among them, 293 were excluded because of inappropriate study design or control samples. Of the remaining 28 relevant articles, 8 articles were excluded for using the same patients. 2 articles were also excluded for their controls in concordance with Hardy-Weinberg equilibrium (HWE). With strict including criteria, the final pool of eligible articles consisted of 18 articles involving a total of 10,619 patients with NHL and 12,977 healthy controls. Because of the large sample size, Caucasians and Asians were considered as population stratification in this meta-analysis. Table 1 shows characteristics of eligible articles including ethnicity, genotyping method, number of cases and controls and NHL pathological types. In fact, at the primary data extraction, allele frequencies in all controls of one study, which performed by Skibola, did not fulfill HWE [25]. This study is a meta-and pooled analysis adding more genotyping data to the initial pooled report [26] to confirm the association between TNF/LTA polymorphism and NHL risk in Caucasian and Asian populations. For the new subjects (including Caucasians and Asians) not included in the initial report (all were Caucasians) [26], allele frequency of TNF-α-308G>A in controls met HWE (p = 0.916). But TNF-α-308G>A in all controls in the initial report was not consistent with HWE (p = 0.0007). We analyzed the initial report composed of 8 subgroups comprehensively [26]. Finally, we excluded data of EPILYMPH-Spain, University of California San Francisco and the NCI-SEER Seattle subgroup, in which controls did not fulfill HWE, and extracted data successfully. Since Skibola et al. [25] conducted this large pooled analysis in TNF polymorphism on NHL risk in Caucasians and Asians, we separated this paper into two studies according to population. In addition, 13 studies were conducted on Caucasians [16,[27][28][29][30][31][32][33][34][35][36][37][38], and 4 were on Asians [39][40][41][42]. Several genotyping methods were used, including allelic specific polymerase chain reaction (ASPCR), polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP), polymerase chain reaction-solid-phase minisequencing (PCR-SPM), polymerase chain reaction-ligation detection reaction (PCR-LDR), TaqMan, Sequenom and sequencing. Abbreviations: NHL, non-Hodgkin lymphomas; ASPCR, allelic specific polymerase chain reaction; PCR-RFLP, polymerase chain reaction-restriction fragment length polymorphism; PCR-SPM, polymerase chain reaction-solid-phase minisequencing; PCR-LDR, polymerase chain reaction-ligation detection reaction; HCL, hairy cell leukemias; FL, follicular lymphomas; CLL, chronic lymphocytic leukemias; MALT, mucosal-associated lymphomas; BCL, B-cell lymphomas; TCL, T-cell lymphomas.

Quantitative Synthesis
Based on a large pooled sample size, we analyzed TNF-α-308G>A polymorphism effects on risks of NHL, BCL, TCL and subtypes (DLBCL, FL, CLL/SLL, MCL, MALT, PTCL and NK/TCL) in additive model (A vs. G) which stratified by ethnicity (Caucasians and Asians). Results of meta-analysis and primary data extracted from studies are listed in Tables 2 and 3.
To evaluate the influence of each study on the pooled ORs in two subgroups, we deleted single study at a time to recalculate the influence of individual study for the outcome of the meta-analysis. The pooled ORs were stable and in an effective interval with statistical significant though the fixed-effect in additive model estimating before or after any single study deleted in each group (data not shown). These indicated that the results of this meta-analysis were reliable and had not been overly influenced by any one of studies.
We performed Begg's funnel plot and Egger's test to evaluate the publication bias of all included studies. Figure 2 shows no evidence of obvious asymmetry in overall analysis for TNF-α-308G>A polymorphism in additive model (p Begg's = 0.529). Egger's test also suggested no significant publication bias existed in this meta-analysis (additive model, p = 0.780).  Begg's funnel plot for publication bias on the association between TNF-α-308G>A polymorphism and NHL risk in additive model.

TNF-α-308G>A and B-or T-CL
Thirteen studies comprising a total of 20,064 participants (8092 cases with BCL and 11,972 controls) and 4 studies including 11,832 participants (1100 cases with TCL and 10,732 controls) were analyzed for an association between TNF-α-308G>A polymorphism and BCL or TCL risk.
No associations were found in TNF-α-308G>A polymorphism with FL, CLL/SLL, MCL, MALT, PTCL and NK/TCL in overall and each ethnic subgroup.

Discussion
In the pooled analysis of 18 articles, we found TNF-α-308G>A polymorphism to be significantly associated with NHL risk in Caucasians and Asians. We provided evidence that subjects with TNF-α-308A allele had an increased risk of NHL in Caucasians, and had a decreased risk in Asians. Similar results were confirmed in analyses of BCL and DLBCL. Further, the TNF-α-308A allele was positively associated with risks of TCL in Caucasians. Our study highlights the effect of TNF-α gene polymorphism on risks of NHL and its subtypes in different populations. These findings indicate a potential connection between constitutively higher TNF-α expression and pathogenesis of NHL.
TNF-α is a transmembrane protein and mainly produced by macrophages and is expressed at low levels in a wide variety of cells. TNF-α mediates its effects through TNF-α receptor 1 and 2 (TNFR1 and TNFR2) by ligand passing and signal transduction. TNFR1 has a death domain that could interact with TNF-α receptor-associated death domain (TRADD), sequentially recruiting proteins to induce caspase-3 activation for apoptosis. TRADD could also bind TNF receptor-associated factor 2 (TRAF2) to recruits proteins activating IKK, GCK, and RIP, which finally leads to the NF-κB, JNK, and MAPK pathway activation for anti-apoptosis and cell survival. Although TNFR2 lacks the death domain, it could also bind TRAF2 to active an anti-apoptosis pathway [13,43]. Aberrant NF-κB activation is a hallmark of several lymphomas for promoting continuous lymphocyte proliferation, which is also directly linked to disease promotion [44,45]. When cells are exposed to TNF-α, NF-κB pathway activation leads to the expression of many genes to cause chronic inflammation, which stimulates tumor growth. Dysregulated TNF-α contributes directly to the transformed state in many cancers, especially those of BCL [46]. Collectively, TNF-α acting as an immunoregulatory cytokine builds a bridge between inflammation and cancer by activating many biological pathways including the nuclear factor-κB (NF-κB) pathway in promoting cell proliferation, survival, transformation, invasion and angiogenesis.
Previous studies suggest higher expression of TNF-α is associated with NHL risk at the time of diagnosis [14][15][16][17]. These studies do not contradict the results of TNF-α-308A allele inducing higher constitutively TNF-α expression associated with decreased NHL risk in Asians. With a heterogeneous malignancy and population diversity, we believe the level of constitutively TNF-α expression must play a different vital role at the step of NHL initiation in Caucasians and Asians, although the reasons for this have not been understood. Once a tumor forms, it secretes TNF-α to promote its survival, proliferation and metastasis. A subtype of TCL failing to express TNF-α and frequently with the TNF-α gene promoter methylated [47] indicates that epigenetic changes might also influence NHL susceptibility together. Environmental, occupational exposure and pathogenic agent infection (such as Epstein-Barr virus and human T-cell leukemia virus-1) are the well-known risk factors for NHL [48,49]. Therefore, genetic, epigenetic, tumor microenvironment, environment and their interaction could together contribute to NHL progression. Few studies about this kind of interaction relative to NHL susceptibility have been published. Due to insufficient data, our meta-analysis did not combine the effects of these factors in an association analysis between genetic variation and NHL risk. Much more precise investigations should be performed to clarify the true association of these types of interactions with polymorphism and NHL.
TNF-α and LT-α gene lie in the major histocompatibility complex class III, telemetric to the class II and centrometric to class I gene. Therefore, TNF-α being in linkage disequilibrium (LD) with these genes may also be linked to another region, haplotype or extended, that can influence NHL development [50]. This meta-analysis only evaluated one SNP in the TNF-α gene, though it was not possible to analyze haplotypes with the present data. Further studies will be needed to pool data and analyze whether haplotypes comprising TNF-α-308G>A and other SNPs are linked to NHL risk, and clarify their function concomitantly.
Because of relatively low incidence of NHL, sample sizes of some studies included in this analysis are very small. The major strength of this study is the larger pooled sample size involving a total of 10,619 patients with NHL and 12,977 healthy controls for association study, which would largely minimize the possibility of chance findings. In addition, patients in our study were all Caucasians or Asians, which would exclude the biased results due to population stratification.
In conclusion, we performed the first comprehensive meta-analysis involving 10,619 patients with NHL and 12,977 controls from 18 articles to evaluate the association between TNF-α-308G>A polymorphism and NHL risk. Our study showed that TNF-α-308G>A SNP in the promoter region of TNF-α gene is associated with NHL risk. In addition, TNF-α-308A increases risks of NHL, BCL, TCL and DLBCL in the Caucasian population; however, interestingly, it reduces risks of NHL, BCL and DLBCL in the Asian population. This association might be mediated by constitutive changes of TNF-α expression in individuals carrying the -308A allele, to induce inflammatory responses or the alternative pathway which be involved in NHL initiation and progression. Our meta-analysis emphasizes that genetic variation plays a crucial role in cancer; theTNF-α-308G>A polymorphism might play a role in a specific subtype of NHL and its importance varies in different populations. Further studies should focus on the function of how variants affect NHL in different populations and the elucidation of the pathway involved which may eventually lead to a better understanding of tumorigenesis and contribute to the prevention of NHL.

Publication Search
We carried out a search in three electronic databases PubMed, Embase and Cochrane Library to find relevant publications up to November 2013, using key words related to the TNF-α gene polymorphism in combination with various NHL subtypes [51]. The search was limited to studies that had been conducted on human subjects and without language restriction. Reference lists of the retrieved articles, reviews and editorials were also screened to find all additional eligible studies.

Inclusion Criteria
Selection of studies had to meet the following criteria: (1) case-control studies, family or sibling pairs studies were excluded; (2) published in English; (3) subjects were limited to adult and without autoimmune diseases, studies with children were also excluded; (4) DNA was extracted from peripheral blood leukocytes; (5) study described the association between TNF-α-308 polymorphism and NHL risk; (6) sufficient data for estimating odds ratio (OR) and its corresponding 95% confidence interval (95% CI); (7) control group fulfilled HWE. When the same subject group occurred in more than one study, only the complete study was chosen to be included in this meta-analysis.

Data Extraction
An initial screening of title and abstract was performed for the first step, followed by further screening based on full-text review. Information was independently extracted from all eligible publications by two investigators (K.Z. and J.D.), including the first author, publication year, ethnicity, sample size, genotyping method, the number of each genotype in cases and controls. For studies with subjects of different ethnic groups and had sufficient information, we extracted data separately for each ethnicity. Disagreements were resolved through discussion.

Statistical Analysis
We assessed the association between TNF-α-308G>A polymorphism and NHL risk by crude ORs and 95% CIs in an additive model. Heterogeneity among studies was examined with I 2 statistics. In this meta-analysis, I 2 > 50% was defined as heterogeneity. Fixed-effect model (Mantel-Haenszel method) was used to evaluate inter-study heterogeneity. If heterogeneity existed, random-effect model (DerSimonian-Laird method) was used. Z test was used to determine the pooled OR and 95% CI. Analyses were also conducted on the subgroups of studies based on ethnicity. The potential influence of publication bias was assessed using Begg's funnel plot and Egger's linear regression test [52,53]. To evaluate the effect of one single study on overall risk of NHL, sensitivity analyses by excluding every study and recalculating ORs and 95% CI were conducted [54]. HWE in controls of each study was examined by the Pearson's goodness-of-fit χ 2 test. All statistical tests were carried out with SPSS 16.0 (SPSS Inc., Chicago, IL, USA) and Stata 12.0 (StataCorp, College Station, TX, USA). A 2-tailed p < 0.05 was considered as statistical significance.

Conclusions
We performed the first comprehensive meta-analysis involving 10,619 patients with NHL and 12,977 controls from 18 articles to evaluate the association between TNF-α-308G>A polymorphism and NHL risk. Our study showed that TNF-α-308G>A SNP in the promoter region of TNF-α gene is associated with NHL risk. In addition, TNF-α-308A increases risks of NHL, BCL, TCL and DLBCL in the Caucasian population; interestingly, this polymorhism reduces risks of NHL, BCL and DLBCL in the Asian population. This association might be mediated by constitutive changes of TNF-α expression in individuals carrying the -308A allele, to induce inflammatory responses or the alternative pathway which be involved in NHL initiation and progression. Our meta-analysis emphasizes that genetic variation plays a crucial role in cancer; the TNF-α-308G>A polymorphism might play a role in a specific subtype of NHL and its importance may vary in different populations. Further studies should be focused on how variants affect NHL in different populations and the elucidation of the pathways involved which may eventually lead to a better understanding of tumorigenesis and contribute to the prevention of NHL.

Conflicts of Interest
The authors declare no conflict of interest.